Systematic palaeontology of late Miocene lagomorphs from the Aït Kandoula Basin (Morocco)

.


INTRODUCTION
The African lagomorph record includes few genera, particularly concerning the Ochotonidae. Although some genera of Leporidae have been described in North Africa (such as Lepus Linnaeus, 1758, Serengetilagus Dietricht, 1941, Trischizolagus Radulesco & Samson, 1967, only three genera of Ochotonidae have been described in the whole African continent, all being extinct: Alloptox Dawson, 1961, Kenyalagomys MacInnes, 1953(synonym of Austrolagomys Stromer, 1926 and Prolagus Pomel, 1853. The discovery of the genus Prolagus in North Africa ( Fig. 1) has increased understanding of the faunal exchanges between Europe and Africa during the Messinian period. Indeed, it was previously believed to be endemic to Europe, except for some specimens of Prolagus oeningensis König, 1925 andProlagus sorbinii Masini, 1989, which had been uncovered in Turkey, in the Miocene of Pasalar (Sen 1990) and the Pliocene of Develi (Sen et al. 1989), respectively. Moreover, in addition to the Prolagus genus, some European taxa such as the rodents Castillomys Michaux, 1969, Stephanomys Schaub, 1938, Occitanomys Michaux, 1969and Eliomys Wagner, 1940 have been described in several African localities dated from the Messinian to the Pleistocene (Brandy & Jaeger 1980;Coiffait 1991;Benammi et al. 1996;Benammi 1997;Mahboubi 2014). On the other hand, some African taxa like the Camelidae Paracamelus Schlosser, 1903 or the rodents Paraethomys Petter, 1968, Myocricetodon Lavocat, 1952and Debruijnimys Castillo & Agustí, 1996 have been identified in Europe during the same period (Jaeger et al. 1975;Minwer-Barakat et al. 2009;Agustí et al. 2011). It has been posited that these taxa migrated during the Messinian salinity crisis, due to the temporary closure of the Betic and Rifian corridors (López-Martínez 1974). However, the discovery of Prolagus remains in the AF1 level of the Afoud site, dated 6.1 Ma in Benammi et al. 1996, then 6.21 Ma in Gibert et al. 2013, indicates that the faunal exchanges between Europe and Africa began before the Messinian salinity crisis, which only started at 5.97 Ma (Benammi et al. 1996;Benammi 1997). Several phases of successive migration have been identified (Gibert et al. 2013) and with the exception of the migration of Hippopotamidae Hexaprotodon Falconer & Cautley, 1836 (about 6.3 Ma), whose transition from one continent to another was facilitated by its semi-aquatic lifestyle, the oldest migration phase occurred when the Betic Strait closed at around 6.2 Ma (Mahboubi 2014 states that this migration phase included Prolagus, Castillomys, Stephanomys, Paracamelus and Paraethomys). Later, this closure contributed to a perturbation of oceanic circulation involving the Messinian crisis (Pérez-Asencio et al. 2012;Pérez-Asencio et al. 2013).
In 2012, more Prolagus remains were identified along the Afoud section of the Aït Kandoula Basin in the southern High Atlas of Morocco (Fig. 2), described as part of Mahboubi (2014). The section had previously been described by Benammi et al. (1996) and Benammi (1997); it is mostly made up of a succession of clay layers interbedded with limestone, marl, or conglomerates (Fig. 3). A volcanic ash layer located in the upper part of the section has been dated at 5.9 ± 0.5 Ma, which allows us, along with biochronological and magnetostratigraphical data, to give a precise age for each fossiliferous layer (Benammi et al. 1996;Remy & Benammi 2006). Thus, the new fossiliferous levels Afoud 12-1 and Afoud 12-2 are dated to 6.12 Ma and 5.9 Ma, respectively. These levels reveal particularly abundant material, which distinguishes them from all other ochotonid sites in North Africa. The treated material has been assigned to the species Prolagus michauxi López-Martínez, 1975in Mahboubi (2014. Attributions or at least close relations to this species have already been made in many localities from Western Europe and in some Mio-Pliocene North African sites associated with European affinity rodent remains, including for instance Argoub Kemellal-1 (Coiffait 1991) and Aghouri (Benammi 1997, with specimens identified as Prolagus cf. michauxi) in Morocco, or Aïn Guettara (Brandy & Jaeger 1980, with specimens identified as Prolagus aff. michauxi) in Algeria. The present work consists of the description and taxonomic study of the lagomorphs from the AF12-1 and AF12-2 levels, collected in 2012 and subsequently. We hypothesize that only one species of lagomorph crossed the Mediterranean Sea during the Messinian period: P. michauxi. The considerable number of specimens of this taxonomic study allows us to test this hypothesis. A comparison of the material will also be performed with some specimens from other African localities, already attributed to P. michauxi or P. cf michauxi. This comparison will be made to confirm or disconfirm the taxonomic attribution of these other specimens, the mate-rial being scarce in these localities and incorrect attributions to P. michauxi or P. cf michauxi are commonplace even for European Prolagus (Angelone & Sesé 2009).

MATERIAL AND METHODS
The study is based on all the material collected in 2012 as part of the thesis of Mahboubi (2014) in addition to the material found during the subsequent missions in 2015 and 2016. In total, more than a thousand isolated teeth have been collected in addition to rare fragmentary mandibles. The majority of the material comes from the AF12-2 level, while only around thirty teeth are from the AF12-1 level. The material is stored at the Université de Poitiers. The percentages of occurrence of various characters were counted (not taking into account damaged specimens), while the length (L) and width (W) were measured in all teeth when ever possible along with the partial width (AA), hypoflexus depth (PH) and distal hypercone length (TH) for the P4 (measurements introduced in Angelone & Sesé 2009). These dimensions are displayed in Table 1; the detailed measurements and the character states for each specimen can be found in the Supplementary material (Appendices 1; 2). We applied a Shapiro-Wilk test to verify the normal distributions of our dental measurements. The L and W dimensions were exploited to draw some scatter plots that also include data from the literature. The reference measures used for the scatter plots, derived from the literature, are listed in Appendix 3 and pertain to the species P. michauxi, P. sorbinii, Prolagus italicus Angelone, 2008, Prolagus latiuncinatus Angelone & Cermák, 2015and Prolagus bilobus Heller, 1936. The occurrences of characters were used in the morphological comparison with other species and in a multiple correspondence analysis (MCA). These three analyses helped us to determine the number of lagomorph species in the population from Afoud. Only the characters of the p3 were used for the MCA. For all statistical analyses performed, the illustrative variable "wear" was used to describe each tooth from our assemblage, in order to differentiate ontogenic variability and potential inter-specific variability. A tooth was labelled "unworn" if it had rounded and bulb-like cusps with a smooth aspect (sometimes with enamel covering the whole occlusal surface) and/or the absence of some connections between cusps.
The teeth were observed with a binocular magnifier and measurements were made with a Mitutoyo monocular measuroscope.
Occlusal drawings of the specimens were made with a Leica binocular microscope. Statistical analyses were performed with R software version 4.0.0. The nomenclature used to describe the teeth is from López-Martínez (1974) (Fig. 4).
The morphology of the specimens from Afoud was compared with European specimens from the literature; these specimens are listed in Appendix 4. A morphological comparison between the material from Afoud and material from other African sites was also performed to try to confirm the attributions previously made at these sites based on very few specimens (see Appendix 5 for the list of African populations directly observed for this purpose). hypsodont and rootless. The anteroconid in the p3 is isolated in the vast majority of specimens, but an occasional connection to the protoconid or metaconid is possible. A protoconulid may be absent, incipient or well-developed. The centroflexid in p3 can be absent, shallow or deep. The trigonid and talonid of the p4 and m1 form two lobes of similar width. A third lobe, the hypoconulid, is always and exclusively present in m2 and is partly or completely isolated from the hypoconid-entoconid lobe. The m3 is always absent. The P2 has a distinct meso-and paraflexus and the mesial hyperloph can be very short to well-developed. The P3 has a well-developed paraflexus and mesoflexus and a shallow and relatively narrow hypoflexus may be present. The P4 is molariform and usually includes one or more rudimentary enamel fossettes. Enamel fossettes in the M1-M2 may be present or absent and may reduce or become lost with wear. The reduction of fossettes in P4-M2 is more progressive towards the posterior, while the depth of the hypoflexus increases.  The mesial hyperloph is always present, usually reaching the centrocone without exceeding it, even though some specimens have a shorter (10% of the specimens) or longer (20%) hyperloph. Its shape is mostly curved and smooth (80%), but sometimes relatively straight. A lingual fold, more or less marked, is present at the start of the hyperloph for 20% of the specimens. An enamel hiatus at the tip of the hyperloph can appear occasionally (10% of the non-damaged specimens) but is rarely visible in occlusal view. The paraflexus is often curved and sometimes widened at the distal tip (30% of the specimens). The mesoflexus can be curved (66%) or straight; over half of the specimens have a distally widened one. The mesoflexus is almost always longer than the paraflexus. One specimen has an isolated centrocone. The postcone is rarely indented (7% of the specimens).

P3 (Fig. 5L-M)
The lagiloph (the structure joining the lagicone to the centrocone) is usually long but rarely reaches the length of the mesial hyperloph. The hypoflexus depth is variable. Two teeth out of three have an enamel hiatus on the precone. The mesial tip of the postcone is rounded to flat and almost never indented.

P4 (Fig. 5N-O)
The distal lobe is longer than the mesial one; they are both of variable width. When they are not broken, the postolobule is generally longer than the precone. The parafossette is long and curved, while the metafossette is smaller and C-shaped. The hypoflexus accounts, on average, for 40% of the tooth.
Upper molars (Fig. 6C-G) The upper molars are composed of two lobes separated by a deep hypoflexus. Fossettes are variable in size, shape and number: nearly 50% of upper molars do not have a fossette, 25% have only a small fossette, 8% have a long or large fossette and the remainder have two or three fossettes (17%).

p3 (Fig. 5A-F)
None of the p3 have a protolophid connecting the protoconid and the metaconid. The crochet is present in the vast majority of the specimens (about 80%); usually it is quite small. It is located in a central position or is displaced towards the metaconid. The anteroconid is more or less displaced lingually and its shape is variable, from triangular to rounded; more than half have a trapezoidal shape. The metaconid is usually rectangular (75%), otherwise it is fan-shaped or rounded (the latter shape is often found for the small-sized teeth that likely belong to young individuals). A slight indentation is sometimes observed on the anteroconid and on the buccal side of the metaconid. Quite rarely, the metaconid is indented distally (30% of the specimens), because of the growth of the mesoflexid. The metaconid is of the same size or slightly smaller than the anteroconid. The shape of the mesoflexid is variable, as it can be straight, curved, with mesial growth related to the indentation of the metaconid, sometimes with a lingually or buccaly widened apex and it is generally short to medium-size. Three specimens have a mesoflexid closed lingually. The protoconulid is generally of medium to reduced thickness with a variable length. An enamel hiatus is fairly often present on the entoconid (40% of the specimens). Around twenty teeth from AF12-2, probably fairly worn, have a connection between the anteroconid and the protoconulid. By contrast, some teeth of young appearance and often small in size do not have any connection between the protoconid and the hypoconid and/or between the metaconid and the entoconid.
p4 and lower molars (Fig. 6A, B) The lobes of the p4 and m1 are kept together by cement. The m2 are three-lobed.

compArisons
In addition to the primary shearing blade typical of lagomorphs, a secondary shearing blade is present on the posterior lobe of the upper molars, like all Ochotonidae (Von Koenigswald et al. 2010). The two shearing blades are also present on the lower molars and the anterior and posterior lobes on them do not have any lingual connection. The p3 has the characteristic morphology of the post-lower Miocene Prolagus. It is triangular with a centroflexid and there is no connection between the metaconid and the protoconulid. Prolagus also lacks an m3 and has a three-lobed m2; this morphology is unique among the Ochotonidae. The observation of the mandibles found in Afoud clearly shows these characteristics, confirming the attribution of these specimens to the genus Prolagus.
In the fossil record of Prolagus, the presence of a welldeveloped mesial hyperloph on the P2 is known to be found occasionally on specimens from the middle Miocene, with Prolagus major López-Martínez, 1977(Hordijk 2010, but it only became very frequent at the beginning of the MN13 (Angelone 2007). Thus, most of the Messinian and post-Messinian species have a well-developed hyperloph on all of their specimens. The absence of the protolophid connecting the metaconid to the protoconulid in the p3 as well as the presence of a fairly well-developed hyperloph on all the P2 render the material from Afoud different from all Lower and middle Miocene species as well as from the more recent species Prolagus crusafonti López-Martínez, 1975, Prolagus ibericus López-Martínez, 1975 and Prolagus pannonicus Angelone & Cermák, 2015. The insular species Prolagus apricenicus Mazza, 1987, Prolagus imperialis Mazza, 1987, Prolagus sardus (Wagner, 1832 and to a lesser extent Prolagus figaro López-Martínez, 1975 have a strong indentation on the postcone of the P2 which is not present on the material from Afoud. Prolagus figaro also has a p3 with particularly crenulated enamel, distinguishing it even further from the Afoud specimens. Prolagus caucasicus Averianov & Tesakov, 1998 has a p3 that is much longer than it is broad and a P3 of much larger size than the Afoud specimens. Prolagus calpensis Major, 1905 andProlagus depereti López-Martínez, 1975 do not have any crochet on the p3 (or it is reduced), precluding the Afoud specimens from belonging to these species. The crochet of P. michauxi is frequent and of variable size. The crochet of P. sorbinii is often reduced, but it can be bigger or quite variable in some populations, like those from the Italian Messinian localities of Brisighella 25 and Ciabòt Cagna. The crochet of P. italicus is often present (50% at Montagnola Senese) and is of variable size. The crochet of P. bilobus of Gundersheim (Germany, MN15) and P. latiuncinatus of Polgárdi 2 (Hungary, MN13) is very welldeveloped. The fairly small size of the crochet of the specimens from Afoud 12-2 is clearly different from the size of the last two species. In addition, the crochet of P. latiuncinatus has a drop-like shape, which is nearly absent here.
The descriptions of P. italicus from the Italian localities Montagnola Senese (MN17) and Torre di Picchio (possibly MN17) indicate that the anteroconid and the metaconid sometimes have a very indented shape, whereas the cusps of p3 from Afoud are mostly non-indented. In the latter case the indentation is almost always less marked than in the Italian specimens. Prolagus italicus also has an anteroconid that is larger than the metaconid, which is rare in the Afoud specimens. Prolagus latiuncinatus can also have an anteroconid that is larger than the metaconid. In addition, these two species differ from the specimens from Afoud because of the morphology of their mesoflexid, which is not as variable as that of the Moroccan specimens. Prolagus latiuncinatus, P. italicus and P. bilobus do not have any enamel hiatus in the entoconid and in this regard differ even more from the Afoud specimens. The protoconulid of the Moroccan specimens is on average much smaller than those of P. bilobus or P. italicus. It also differs from that of P. michauxi of Sète (France, MN15) but resembles the one from Kessani (Greece, MN13/14).
Nearly half of the upper molars do not have fossettes (or are too damaged to allow observation of any). Those fossettes that do exist can be of very diverse size and shape (Fig. 6  have at least two fossettes, with three of them (<2% of the upper molars) having three fossettes or more. The number, shape and size of these fossettes distinguish the specimens of Afoud 12 from all other known species, with the possible exception of P. italicus. P. sorbinii from Monte Castellaro and P. aff. sorbinii from Case Inferno are slightly larger than our specimens. The other populations of P. sorbinii and P. cf. sorbinii (various localities), P. bilobus (Tanatary, Tatareshty), P. michauxi (Kessani) and P. latiuncinatus (Polgárdi 2) are, on the whole, included in the variation of the specimens from the present work. The P3 (Fig. 7C) of P. cf. sorbinii from Velona are clearly smaller than the Afoud specimens. The specimens from Monte Castellaro, Montagnola Senese and Kessani are slightly larger than our specimens.
The species P. michauxi and P. sorbinii seem to be the species with the closest morphologies and dimensions to the Afoud specimens. Due to the considerable intraspecific variability between the two former species, a population-level comparison is needed to assign our material to a species in order to discuss the possible Europe-Africa migration patterns.
Prolagus michauxi from Sète has a much larger crochet and protoconulid than is true of our specimens; furthermore, the enamel hiatus on the mesial hyperloph of the P2 is much more frequent at Sète. There is no information about the variability of the metaconid or the mesoflexid of the p3, but the mesoflexid illustrated in López-Martínez & Thaler (1975: fig. 4, pl. II) is much larger than that of our specimens. There is also no precision about the possible presence of a hiatus on the entoconid. The anteroconid has a triangular shape, which can also be found in Afoud without it being predominant.
Prolagus michauxi from Granada (Spain, Messinian) can be studied here only through a single illustration of the p3 (originally from López-Martínez 1989, appearing in Angelone 2008b: fig. 5, for example). However, the morphology of this p3 is very similar to that of our specimens. The crochet is present without being very large, the metaconid is rectangular, the mesoflexid has a medium size with a mesial growth and the protoconulid is not excessively thick. Nevertheless, the variability of its characters is not known because we did not have access to descriptions, so we cannot safely assign our specimens to the population of Granada.
Prolagus michauxi from Kessani has been described as having a p3 with a large, asymmetric and rectangular to oval-shaped metaconid, while the lagiloph of the P3 is long without reaching the outline of the tooth. There is no hiatus on the P2, which does not significantly differ from our specimens where the hiatus is rare. The mesoflexid of the p3 has been described as being long, but the representations from Vasileiadou et al. (2012: fig. 3f ) show mesoflexids with a size that is included within the variation of our specimens. The P3 illustrated in the same study (Vasileiadou et al. 2012: fig. 3c) is quite similar to those from Afoud. However the p3 of this population differs from our material owing to its more rounded anteroconid; moreover, its crochet is at least "moderately long", which suggests that it is never absent or reduced; the length of the P3 is also greater at Kessani. The variability of the mesoflexid of the p3 is not mentioned. Vasileiadou et al. (2012) note that Syrides et al. (1997) also found material concerning Prolagus in Kessani, but attributed it to P. aff. michauxi. A drawing of a p3 from the lat-ter study can be found in Angelone (2008a: fig. 1), which is very similar to some specimens assigned to P. michauxi from the same site.
Prolagus michauxi from Silata (Greece, MN13/14) has a rounded metaconid (in contrast to our specimens) and the entoconid is very thin. The hiatus on the P2 is also much more common than in the Afoud specimens. We can note the variability of the crochet and the anteroconid.
The specimens from Brisighella (Italy, Messinian) described as P. sorbinii have fairly varied morphologies, depending on the karst fissures from which they were extracted. The p3 from Brisighella 25 (Angelone 2008b: fig. 5) are very similar to the specimens from the Afoud locality, as they differ only in the size of the crochet, this being larger in Brisighella. On the other hand, the specimens from Brisighella 6 (Angelone 2007: fig. 4e) have a triangular metaconid and a much more slender shape, while their crochet is relatively small like on the Afoud specimens. Some specimens from Brisighella (Angelone 2007) have an indentation on the postcone of the P2, very rare in the Afoud specimens, but the fissure from where they came has not been specified.
Prolagus sorbinii from Monte Castellaro (Italy, MN13) has slightly larger p3, P2, P3 and P4 than our specimens (Fig. 7). The protoconulid on the p3 is extremely well-developed -especially in comparison with the very reduced protoconid-and the protoisthmus is often absent. The upper molars do not have any fossettes, or they are reduced.
The Prolagus from Ciabòt Cagna (Italy, MN13) were first described in Cavallo et al. (1993) and more information was subsequently provided by Angelone (2007). However, Angelone & Cavallo (2010) have deemed this description valid for the "radical" side of the tooth rather than the occlusal one, which is broken. Thus, these specimens, first classified as P. michauxi and then P. sorbinii after the revision, cannot be assigned to the Moroccan specimens.
The p3 of P. sorbinii from Verduno (Italy, Messinian) illustrated in Colombero et al. (2014: fig. 12.1) does not seem very different from our specimens, except that the global shape is quite slender and the metaconid is considered triangular. There is no information about the variability of the characters.
The illustrated specimens of P. sorbinii from Develi (Turkey, MN14) in Sen et al. (1989: figs m-q) include a P2 with a very large and protruding centrocone (very different from the Afoud specimens) and a p3 with a rounded anteroconid and metaconid.
The p3 of P. sorbinii from Grebeniki 2 and Novaya Andriyashevka (Ukraine and Moldova, MN14) have a rounded anteroconid and a short mesoflexid and their crochet is often absent, while the P3 seems more variable than our specimens.
Prolagus aff. sorbinii from Arcille (Italy, MN16a) has a very well-developed protoconulid and a buccal indentation on the anteroconid. In addition, the paraflexus and the mesoflexus of the P2 have the same size and shape; they are straight and thin, in contrast to our specimens. An enamel hiatus on the hyperloph is present, but the frequency has not been specified. Conversely, there is no hiatus on the entoconid Late Miocene lagomorphs from Aït Kandoula COMPTES RENDUS PALEVOL • 2022 • 21 (40) of the p3. The crochet has neither been mentioned nor illustrated, while the centroflexid is undulated and curved towards the protoconid, in contrast to the crochet in the Afoud specimens, which is displaced towards the metaconid. The metaconid's variability has not been mentioned, althrough the one illustrated in Angelone & Rook (2012: fig. 3o) is fan-shaped. The size of the p3 is distinctly larger than those of Afoud. Similarly, P. aff. sorbinii from Case Inferno (Italy, Pleistocene) also has a well-developed protoconulid, a buccally indented anteroconid, an entoconid lacking a hiatus and a centroflexid curved towards the buccal side, in addition to having a fairly large p3 (Fig. 7A).   Prolagus cf. sorbinii from Velona (Italy, Messinian), in addition to having a much smaller P3 than the specimens from Afoud (Fig. 7C), has a p3 with a triangular metaconid, an entoconid without a hiatus and a distal and lateral indentation of the anteroconid more frequent than our specimen. The mesial hyperloph of the P2 is very long and is thinner than our specimens, while the paraflexus and the mesoflexus are verticalized, relatively straight and thin.
The illustrations of P. cf. sorbinii from Maramena (Greece, MN13/14), originally from De Bruijn (1995) and appearing also in Angelone (2007: fig. 4f-h), for example, show a p3 with a triangular yet quite rounded metaconid. The illustrated P2 has a particularly straight and thin mesial hyperloph, paraflexus and mesoflexus. These are very different from the specimens from Afoud.
Prolagus cf. sobinii from Borro Strolla (Italy, Mio-Pliocene boundary) has a metaconid that is very different from the rectangular one from Afoud and the representation in Angelone & Rook (2012: fig. 3e-f ) shows a very long protoconulid; the shape of the p3 seems more slender than in Afoud. The variability of the mesoflexid has not been mentioned; the only morphology evoked is a large and V-shaped one. The lagiloph on the P3 is shorter than those of our specimens.
Therefore, the populations that are closest to our specimens are P. michauxi from Granada and Kessani and P. sorbinii from Brisighella 25 as well as maybe Verduno. Brisighella 25 is only one of the many karst fissures from the Brisighella locality corresponding to the Monticino Quarry. The specimens from this fissure are very different from those of many other karst fissures, even though they have been described as the same species. The filling of these fissures can be of different ages, so it is necessary to study them independently; as a result, the difference between P. sorbinii from Brisighella 25 and those from the other fissures do not provide a sufficient argument to exclude the specimens from Brisighella 25 from our reasoning. However, the large size of the crochet, which is their only clear difference from our material, cannot be observed in the p3 from Granada, this tooth in fact manifesting no difference from the p3 from Afoud. Prolagus michauxi from Granada looks more like our specimens than P. sorbinii from Brisighella 25, but the fact that the tooth from Granada is only studied here through a unique representation, without description, does not allow us to come to an immediate conclusion, given that we do not know the variability of the population. The population from Verduno has a slender p3 and a triangular metaconid, while the one from Kessani has a slightly rounded anteroconid and a fairly large crochet in addition to a quite long P3. However, the variability of the metaconids found within the Afoud population has only been described in the specimens from Kessani.
The great resemblance between the tooth from Granada and our material, combined with the variability of the metaconid in the population from Kessani, ultimately allows us to assign the specimens from the Afoud locality to the species P. michauxi.
Prolagus michauxi and P. sorbinii are dentally very similar; in fact, a morphometric (complex discrete fourier transform, CDFT) analysis performed by Angelone (2008a) was unable to distinguish these two species. In this study, it even turned out that the specimen assigned to P. aff. michauxi in the locality of Kessani, used for the comparison of our material, was closer to P. italicus than P. michauxi, despite the resemblance between P. aff. michauxi and the P. michauxi from Kessani. However, these species can still be distinguished using the morphology of their skull. Indeed, in the emended diagnosis of P. sorbinii, Angelone (2007) stated that the premolar foramen of P. sorbinii is larger than that of P. michauxi, that the incisive foramen is divided in two by a narrowing of the premaxillary bone in P. sorbinii (which is not the case in P. michauxi) and that the muzzle is longer for P. sorbinii. Thus, the attributions realized in the current work must be considered cautiously until some more complete cranial material is found in the Aït Kandoula Basin. Until then, the specimens from Afoud are still attributed to the species P. michauxi.
The measures performed on the p3, P2 and P4 from AF12-1 and AF12-2 ( Fig. 7A-B, D; Table 1) encompass the measures of both African and European P. michauxi and P. cf. michauxi from the literature (the African specimens often being smaller than their European counterparts), which is a further argument concerning the assignment of the previous Moroccan specimens (from Aghouri, Afoud 2 and 8) to the species P. michauxi or P. cf. michauxi. Prolagus sp. from the Voie Ferrée site is included within the variation of the Afoud specimens ( Fig. 7A-B) and may be assignable to the species P. michauxi, but direct study of its morphology is needed first.
The presence of a single Prolagus species from Afoud (excluding the P2 mentioned below) is confirmed by the MCA, where the distribution of individuals in morphological space (Fig. 8) reflects the presence of worn and unworn teeth. It is probable that unworn teeth are recently erupted and are therefore likely to belong to rather young individuals.  The fact that these unworn teeth have a relatively small size (while some of them are not necessarily different in size from the smaller fully adult specimens) is an element in support of their labelling as juvenile. Figure 9 confirms this fact: indeed, the character states causing a separation within the morphological space are the short LENGTH, short WIDTH, absent CONMETENT, sub-round ANTSHP, rounded METSHP and to a lesser extent the absent CONHYPPROT, thin apex MEFLXSHP and thin PROTLTHK. The absence of some connections, the rounding and also the small size are characters typical of rather young individuals, their permanent teeth being present but probably not worn. Given that the separation along the first factorial axis -horizontal-is caused by characters related to ontogeny, this factorial axis should thus represent the ontogeny. The bar plot in Figure 8 highlights the fact that this first factorial axis has an eigenvalue much higher than the others (0.2741, representing 11.60% of the information), even without being extremely high itself, showing that this horizontal separation in the morphological space is the only efficient way of making groups with these individuals. In fact, the other factorial axes did not allow us to identify several groups. Therefore, it is likely that the only differences observed from one individual to another are related to wear and to the phenotypic variation within a species, so it is quite unlikely that several species of Prolagus can be present in this assemblage. Histograms (Figs 10; 11) do not point to the presence of several species in AF12-1 and AF12-2. All histograms with a sufficiently high headcount for the results to be significant display a unimodal distribution. Some dimensions are normally distributed when AF12-1 and AF12-2 are treated separately; in AF12-2 (Fig. 11), this is the case for the length of P2 (w of Shapiro-Wilk = 0.98735, p-value = 0.426) and the widths of P2 (w = 0.97705, p-value = 0.0902) and P3 (w = 0.98637, p-value = 0.2055). All other dimensions from AF12-2 do not follow a normal distribution; but their distribution is still unimodal, so it does not contradict the statement of a single species. It is possible that ontogeny, as highlighted previously, can explain this type of distribution. The cessation or slowing down of growth at sexual maturity can explain the shift of the peak towards the right of the histograms: all sexually mature individuals of this species may have a quite similar and large size (with little variability among individuals), no matter whether they are young adults or very old individuals. Conversely, juvenile individuals of different ages are likely to have very different sizes, which may explain the spread of the lower values. The gradual increase in these dimensions during their lifetime can explain why we do not see two different age groups in the histograms.
The morphometric (Fig. 7) and morphological variability of the dental characters of the AF12-1 level is included in the variability of the AF12-2 variation. In addition, when a dimension follows a normal distribution both in AF12-1 and AF12-2 (as is the case of the length of P2 and the width of P2 and P3), the distribution remains normal when we combine the measurement of the two levels (Fig. 12). This implies that the specimen of the two levels can be assimilated to the same population. The specimens from these two levels are thus assigned to the same species.
The morphology of the specimens from Aghouri and Afoud 2 and 8 (Benammi 1997) is not particularly different from that of the specimens from AF12-1 and AF12-2, hence confirming their attribution as P. michauxi and P. cf. michauxi. Conversely, the Prolagus sp. specimens from the Plio-Pleistocene locality of Ahl al Oughlam differ from our specimens (Sen & Geraads pers. com. 2017). Prolagus sp. (Fig. 5K) description And compArisons A unique P2 (AF12-2-L-453) has an extremely marked crenulation in the mesial hyperloph. Although a slight indentation has been observed in some Prolagus species, especially insular, this type of extreme crenulation has not been seen in any known taxon. It is represented in this assemblage by only one specimen, without any intermediate form between the smooth hyperloph of P. michauxi and the crenulated hyperloph of this P2 (some specimens assigned to P. michauxi do have a very slight indentation, but not marked enough to be related to this P2). It is still possible that it is an aberrant individual of the species P. michauxi, but the lack of documentation about anomalies in Ochotonidae impels us to describe this tooth as Prolagus sp.
Family leporidAe Fischer von Waldheim, 1817 Leporidae gen. et sp. indet. (Fig. 6H) description And compArisons In addition to the thousand teeth assigned to Ochotonidae, three upper molariform teeth, clearly typical of the Leporidae, were unearthed. For each of them, the hypoflexus is strongly crenulated and the general size of the tooth is much larger than that of the other Afoud specimens. We did not perform any extensive analysis on these specimens, so we were unable to assign these teeth to a precise genus or species. Examples of Leporidae to which these teeth may be affiliated, with regard only to their palaeogeographical setting, include Trischizolagus, found in some Plio-Pleistocene sites of Morocco (Geraads 1994(Geraads , 2006 (2014) Here we will reuse what was stated in this previous study. It has already been demonstrated that the mammal remains from this level were not subject to post-mortem displacement, according to the absence of polishing traces and the presence of some articulated elements of large mammals. Moreover, we assume that small mammals are very closely related to specific environments, as they are likely to respond very quickly to environmental changes (be it on an ecological timescale or, even more so, on a geological one; the same assumption was made in Van Dam & Weltje 1999, for example). Therefore, the analysis of the Glires and Lipotyphla assemblages from AF12-2 allows the vegetation type and climate relating to this level to be inferred.
Among the micro-mammals from AF12-2 studied in Mahboubi (2014), Ochotonidae (Prolagus) and Muridae are comfortably the predominant groups. Paraethomys represents the largest part of the Muridae, constituting more than 90% of this group. The relative abundances of the Gliridae, Ctenodactylidae, Erinaceidae and Soricidae are fairly low, as they each account for about 1% of the material. Gerbilidae (Protatera, Myocricetodon) and Sciuridae (Atlantoxerus) are a little more present, without being that abundant.
The ecological preferences of Prolagus are not yet well-known. Some species from the late Miocene or the Pliocene were able to live in moist and wooded environments, whereas others

CONCLUSIONS
This study of the dental material from the AF12-1 and AF12-2 levels of the Afoud locality in the Ait Kandoula Basin has confirmed the presence of the species P. michauxi in this site, even if additional cranial material is needed to definitively exclude its belonging to the species P. sorbinii, which has a very similar dental morphology. It is the only known ochotonid to have crossed the Mediterranean Sea during the late Miocene.
The presence of Leporidae at this site, alongside Ochotonidae, is interesting, given that they have not previously been reported in Morocco before the Pliocene. We hope for the discovery of new material to determine the species to which this Leporidae belong.